Knowledge Base Construction Based On Chinese Wikipedia

In recent years,with the continuous development of computer technology,Natural Language Processing has become a popular direction of computer field.On the one hand,the understanding of Natural Language requires a large amount of knowledge,on the other hand,Natural Language Processing also provides a new way for building knowledge base.In this paper,we present a method of extracting data from Wikipedia in the construction of knowledge base,it includes three aspects,the extraction of entity properties and relations,the extraction of entity relations in the attribute values and the graphical storage of knowledge base,the specific research contents are as follows:(1)Extracting the relations and properties of entity from Wikipedia Infobox.We build ontology and determine the data attributes and object attributes of knowledge base through the merging and generalize on synonymy template of Wikipedia Infobox.Data attributes and object attributes are seperated by normalizing the property name,then extract the tuple of property and relation from Infobox.For the largest number of character entities,we custom rules to extract some of the important attributes of the default.(2)Extracting the relation of entities from property of Infobox.The attributes of partial object of Infobox contains multiple entities,we extract entities relations from it.We first do the linking of entity reference and get the Wikipedia item as a feature to effectively improve the performance of entity relationship classification before the identification of entities relations.(3)The storage of knowledge and the test of knowledge base.The paper use graphic database and relational database,Depth relation search on multiple datasets.According to the response time of the two databases and the results of the search,we finally choose to store the knowledge base with the graphics database.The results show that the knowledge base constructed in this paper is easy to use,feasible and robust.